Florida Voter Purge: Analyzing Florida’s 6/19 Letter to the DHS

[This post was updated June 27th, 2012 with information from internal emails]

On June 19th, Ken Detzner, Florida’s Secretary of State, sent a letter to the Department of Homeland Security (this opens in a second window if you wish to read it) secretary Janet Napolitano restating Florida’s desire to get access to the federal SAVE (Systematic Alien Verification for Entitlements) Program database and rebutting the grounds given in the Department of Justice’s letter for not granting access.

His letter tried to achieve two goals.  First, it tries to demonstrate that the process used to identify suspected illegal voters was fair, accurate, and diligent.  Second, it rebuts the DOJ’s statement that the state did not meet the technical requirements necessary to access the federal database and show that the state should be granted access.

Did it achieve its goals? 

The short answer: No and No.

As I mentioned in an earlier post, the argument between the state and federal governments is unusual, for squabbles between governments, because the issue is, at its heart, technical, not legal.  Since this new letter continues the technical argument, it is worthwhile to dissect it to see what it says and whether Florida has made a case that, on the surface, contradicts what the DOJ stated.

Did Florida meet the requirements to access the Federal “SAVE program” database?

In paragraph 3 of the letter, Detzner states that the DOJ is wrong in the substance of its claims by writing:

Unfortunately, the USCIS letter indicates a lack of familiarity with Florida’ s procedures to identify potential non-citizens on its voter registration rolls, FDOS’s intended use of the SAVE Program, and -­ most significantly – the data fields that FDOS offered to provide to USCIS, several months ago, to verify the current legal status of potential non-citizens. As previously stated, the Florida Department of State has access to alien registration numbers and other immigration-related documents for many potential non-citizens on its voter rolls and remains willing to provide this information to USCIS for verification purposes.

As you may recall, the DOJ outlined what it believed to be the two step requirement for use of the SAVE program database: All requests must be accompanied by immigration document numbers and, upon request by the SAVE program, copies of the physical documents must be supplied as well.  The above quoted paragraph leaves open whether all of Florida’s requests will be accompanied with document numbers,  or most, some, a few, or virtually none.  

The answer turns out be “none”, as we can see further in the letter.

In the section headed “Florida Has Offered to Provide the Unique Identifiers Required by the SAVE Program”, we find some real discrepancies in the letter.  The second paragraph in this section says “As explained to USCIS several months ago, the Florida Department of State has access to the alien registration numbers and immigration-related documents  contained in DHSMV’ s database.”  Note that supplying the document numbers from immigration documents is a requirement for using SAVE; it is a system for indicating the validity of such documents, not for validating people’s citizenship status. 

In the third paragraph of this section we see the reason to doubt the state is being honest; the state lists the data they were offering to supply.  Conspicuously missing are any immigration document numbers. None. Nada.  Zero. Zip.   But that’s what’s necessary to access the SAVE database.  Despite having access to this data (so the state claims), the state did not offer to submit it.  Why not?  Because, of course, they really cannot.  They don’t have it for very many records (see the next section of this post for why that is).

Since this letter was published, the Brad Blog (http://www.bradblog.com) released internal email correspondence between Florida’s Department of State and the federal Department of Homeland Security. These emails make clear the state was told, repeatedly, what was needed to access the SAVE database — and make clear that the state could not, and knew it could not meet the requirements.  Unfortunately, the conversation seems to be between a lawyer and a techie, and the lawyer seems to have adopted the “play dumb” approach to dealing with being told what they needed to do.

Which leads us to the last paragraph in this section. It reads, in its entirety,

FDOS has gone well beyond offering only “name and/or date of birth” to USCIS for verification of the legal status of potential non-citizens. Indeed, FDOS has offered to provide the very information the USCIS letter now states is sufficient.

This is simply not true.  The information the letter asks for are the immigration document #s, and in the third paragraph the state says they did not supply it.  Going well beyond name and data of birth does not help.  It’s like answering a math problem wrong on a quiz but doodling a few other, unrelated, math problems on the test to try to get extra credit.

The state does not justify its argument by first trying to bamboozle the reader and then admitting, by omission, that it did not actually do what it needed to do to access the federal database.  That the state admits to this problem in their emails and yet promotes the idea that they were frozen out by Obama’s DHS is just a pathetic attempt to play the victim.

Was the list of suspected voters generated with proper care and diligence?

There’s another set of issues the letter raises — around how the prospective voter purge list was created.  The state wants us to think that they were very careful to only find voters who they could reasonably suspect of being non-citizens.  But as we can discover from the inconsistencies in the actual list that was generated, that is very unlikely to be true.

Paragraph 5, in part, says that:

Florida’ s Department of Highway Safety and Motor Vehicles (DHSMV) maintains a record of citizenship/immigration-status information for individuals who have obtained a state driver license or identification card. FDOS believed DHSMV’ s legal-status information could be helpful in identifying non-citizens who have registered to vote.

This identifies the source of information that they propose to submit to the SAVE program: The database of people with Florida driver’s licenses or identity cards.  But “maintains a record” does not mean, in fact, that there is anything in that record.  A statement that read “a complete and accurate record of … for all individuals” would say that.  But what they said is no different than saying “I have a column in a spreadsheet called ‘status’.”  They didn’t say if there’s anything in it.

Can we estimate what proportion of driver records might have this information, the record of citizenship status?  Florida first instituted a requirement for providing this information on 1/1/2010.  Florida driver’s licenses are good for 8 years.  The timing of the analysis is not known, but appears to be between Spring 2011 (the date mentioned elsewhere in paragraph 5) and October 2011 (the date mentioned in the DOJ letter).  Let’s just call it 1/1/2012 conservatively.  That means that, at first blush, only 1/4rd of all new drivers licenses were subject to this requirement.

That number is very high, however, as the requirement to supply these documents only apply if you renew your license in person and not by mail.  I renewed my driver’s license the last time in person, but I suspect I am in a very small minority.  Nobody else I know  renewed in person; they all renewed by mail.  Let’s say, conservatively, that people renew in person every other license.  That means that out of the 16 years of that people potentially could have visited the license bureau, only 2 were covered by the new requirement, or 1/8th.  And for “illegal” voters and drivers, we might assume they would have avoided that in person visit in any way possible, including by letting their driver’s licenses expire.  No matter, there is no need for precise accuracy here, so we can guesstimate that 10% of all voters who have driver’s licenses (and many may not) were covered by the prove-you-are-here-legally requirement.

That means that about 90% of the records in the database have nothing that indicates the citizenship status of a driver.

What about social security numbers? In years past, the requirement for a social security number was not terribly strong, not well enforced, and subject to fraud and data entry error.  Further, having a social security number means you have permission to work in this country, not that you are a citizen.  It’s not useful for rooting out green card holders who are voting.

Before we move on to paragraph 6 of the document, let’s review what is required to register to vote in Florida: you need to provide either a Florida driver’s license/identity card number OR the last 4 digits of your social security number.  The last part is important, as SSNs are 9 digit numbers.  In addition, many people are exempt from these requirements, such as people 65 or older, deployed military, disabled, etc. Furthermore, if a voter is required to provide ID but does not provide a license/ID # or last 4 of SSN, they will be registered nonetheless with a proviso that they are required to provide identification prior to voting.  There are many ways to be a citizen who is a registered voter with none of this information.  And it is likely that if you were not a citizen but had a strange desire to register to vote, you would probably make up a four digit number and call it a day.

Paragraphs 6  of the letter describes the process used to match the voter roll with the driver’s license database, DAVID:

FDOS worked with DHSMV to identify potential non-citizens who were registered to vote by identifying records common to DHSMV’s Drivers and Vehicle Information Database (DAVID) and the Florida Voter Registration System (FVRS). The matching criteria used for this initial, automated match were

(1) exact match on a driver’s license number;
(2) exact match on a nine-digit social security number; or
(3) exact match on a generated driver’s license number (derived from name (first and last), date of birth, and gender as contained in FVRS)

with existing driver’s licenses in DAVID. This final matching criterion was used because there are registered voters for whom there is no driver’s license number or social security number, as they registered prior to that identification requirement.

 Right away, there are some red flags this paragraph raises.  For match #1, of course some voters will not have supplied a driver’s license or ID # for whatever reason.  For match #2, this will never, ever occur because the voter registration forms do not ask for a nine-digit social security numberI have no idea why the letter was written to say this, as it is clearly inaccurate.  If this is intentional, then it represents an attempt to mislead.  If it is unintentional, then it represents a surprising lack of understanding by the head of Florida’s department of state of how voter registration works.  Given that other parts of the letter are counter-factual, I have to assume this was thrown in to make the state look good even if it is misleading.

For match #3, we have to look at how Florida driver’s license numbers are generated.  If you live in Florida, pull out your driver’s license now and go to this website (not mine) and enter your personal information (OK, you have to trust them, so you can skip it if you wish). You’ll see that it generated almost your correct driver’s license number.  That’s because everything up to the last digit of your license number is generated from your first name, middle initial, last name, sex, and birth date.  This website explains it in complete detail, if you like.  The last digit is used because multiple people can end up with the same computed driver’s license number and so is used to differentiate between them.

If you take the information on the voter registration card and it has been entered correctly into the voter rolls (no mistakes on the part of the person filling out the form or typos by the data entry operator) and the person indicated their sex (this part of the voter registration form is optional) then you can generate a good guess as to the person’s driver’s license number — by adding “-0” to the end of the number, assuming that there are very few -1s and above.  So this will work more often than not when the person has a Florida driver’s license and the information on the voter rolls matches whats in the driver’s license exactly (voter Bob Smith would not match driver Robert Smith at all, even if they are the same person).

Again, because of the uselessness of trying to match on a social security number, the matches will be based only on a driver’s license number, either guessed at or supplied by the voter.

Paragraph 7 explains what happened next:

The return file only contained information on automated matches in which the DAVID system indicated that the person was a non-citizen based on documents produced to obtain a driver’s license or state ID card. DHSMV’s automated match process was able to identify roughly 11 million common records, of which approximately 180,000 were identified by DHSMV as potential non-citizens.

Let’s understand what this means.  First, as May 2012, according to the Florida Division of Elections, there are 11,330,015 registered voters in Florida.  The “roughly 11 million common records” means a match rate is in the worst case 93% (10,500,001 rounded up to 11,000,00) and in the best case 100%.  Even 93% a suspiciously high match rate unless an incredible amount of work was taken, beforehand, to clean up both databases.  Which does not appear to be the case.

The letter says the ~180,00 records were those where the DAVID system indicated that the person was a non-citizen.  “Indicated” is a very interesting choice of words.  It does not mean much.  It could mean “knows for a fact that at the time of getting a driver’s license the person was known to not be a US citizen”.  But if that were the case, then it could not have flagged the 91-year old, US born, WWII hero who was covered by the press.  

As pointed out in this article, both the voter roll database and the driver’s license database are known to have errors, and that comparing the two would “compound the errors”. The  article goes on to say that the former Secretary of State Kurt Browning first generated a list of 182,000 potential illegal voters, which was “too high”, and refined the search criteria to produce a list of 25,000 potentials, which was still “Too high for Browning.”  He said “We didn’t have our I’s dotted and T’s crossed when I was there.”  

However, as you can see from the current secretary of state’s letter, they didn’t even start with the “too high” 25,000 voter list, they went back to the 182,000 potential voter list — they cast the widest net possible.

What would that wide net look like — what would “indicated” really mean? At that, we can only guess. Unfortunately, the state has indicated they won’t release the full list for privacy concerns, so we can’t know for sure. Certainly if the driver’s license information included some sort of immigration status that was not a citizenship (green card #, etc.) that is a possibility.  Also, if the mailing address was not in Florida, that might suggest that the voter was not a Florida resident and therefore not a legal voter here: a snow bird, perhaps, or maybe a resident of Puerto Rico, or someone who has moved out of state and gotten a driver’s license in their new home state.  Remember, non-residency is reason to purge somebody from the rolls just as much as non-citizenship. Finally, if the driver’s license information is old (an expired license from years ago where the driver is, say, a 91 year old WWII vet who can no longer drive), there is going to be just no information at all about the person.  Does that “indicate” an illegal voter? Maybe — it looks similar to someone who is either dead or moved out of state. But we don’t know because the letter is unclear.

Paragraph 8 discusses what happened with those results:

Following the automated match process, the Florida Department of State manually reviews each automated match. The two-step process involves a verification of both identity and legal status. First, the Florida Department of State verifies that the registered voter is the same person as the driver licensee by reviewing the name and date of birth under the driver’s license and/or social security number, along with a review of other common fields such as address and signature comparison. Identification must be verified on a minimum of three secondary match criteria (first/last name, date of birth, and driver’s license number and/or social security number).

This says the Florida department of state had someone look at each of the 182,000 matches.  That is a massive undertaking. Did they just look at the (error-prone, per Browning) data that the state had keyed in the computer systems, or did they review the actual forms filled out by the voters and drivers? Did they do this very carefully? Or was it sloppy with the belief they could let the county supervisors of elections sort it out? (In the botched felon purge of 1999/2000, there was an expectation that the counties would sort out any problems). The county supervisors have indicated that the lists are low quality, so I suspect the manual scrutiny was not well done.  In addition, remember that they do not say that they limited the search to current licenses.  Data from a 1979 driver’s license is going to be pretty shaky.  

Also remember, there is no  full social security number on the voter form.  There is no need to supply even the last 4 digits if you give a driver’s license.  How many people entered the first 4 digits of their social instead?  There’s lots of reasons to doubt this field is useful very often. The SSN match is going to be imprecise if you have any information at all.

Note that the letter does not say how many people passed the manual step.  We don’t know.  Out of the 182,000 — how many remained?  We know they sent off 2700 to the counties — they must have worried that the output was garbage and so wanted to do a test.  And, of course, fear of garbage data was their motivation to try to get the feds to clean up their list for them.

In paragraph 9, they state that when they have a match, they review the information to verify that the information in DAVID “indicates ‘non-citizen'”.  Again, how did the database positively indicate that the WWII hero was non-citizen?  Either the process is terribly flawed, or they marked anyone whose record was not “REAL ID compliant” (described later in the paragraph) as a “potential ineligibility”.  Or, anyone who they could not find in the driver’s database was also marked as potentially ineligible.

One way or another, there are a lot of false positives, and yet the description of the process in the letter would lead the reader to believe that only when there is reasonable doubt of a voter’s status did it produce a hit.  Either they are intentionally misleading us about the process, or it is so fatally flawed as to be useless.

In paragraph 10, they said only names that passed all these tests would be forwarded to the counties.  But the names that passed the test clearly contain many people who do not meet the criteria.  So much so the supervisors of elections are refusing to process the lists.

If the state had followed the process carefully, and the data in the two databases was reliable, they would not have found so many legitimate citizens to accuse.  This means that either the process described in this letter is not accurate or that the data in these systems is not of very high quality, or both.

When you consider that the state lied about meeting the requirements of the SAVE database and the state describes the purge process as being diligent when the results are sloppy, it is very clear that the state is looking to score political points instead of being focused on maintaining a clean voter roll.  When I see tea-party organizations like “True the Vote” spending time and effort to get states to purge voter rolls and institute discriminatory Voter ID rules, I can only conclude that Rick Scott is looking to play to his base — playing with the mechanism of our democratic government like a cat plays with a mouse.